AI applied to Finance work.
Choose a useful task, organise the data, verify the output and retain responsibility for the decision. My use cases start with financial production requirements.
Business experience, supported by AI.
During an assignment for an investment management software provider, the work included reconciliations and controls assisted by the solution’s AI. In my own practice, I develop agent workflows for consulting and publishing.
Reconciliations & controls
Compare sources, isolate exceptions and prepare analysis. Matching rules and amounts remain calculable and verifiable.
Production documentation
Draft procedures, test scenarios and briefing notes from identified sources, with business review.
Agents for consulting and publishing
I use Claude and GPT agents to prepare and review documents, organise research and automate recurring processing.
Scoping & team training
Define an initial scope, permitted data, acceptance criteria and responsibilities. Train teams on their own cases.
An example you can try.
Reconciliation, tolerance and review suggestions: each step is visible.
Try a position reconciliation
Entirely fictional data. The calculations and comments below use deterministic rules, executed in your browser.
| Fictional instrument | BO / custodian quantity | BO / custodian price | Check |
|---|
The comment suggests a check; it does not establish the cause of an exception. In an AI project, this step can prepare human review using authorised sources.
Scope your first use case.
The member template helps describe the task, data, controls and operating requirements.
Questions to resolve at scoping
- What specific work needs to be produced?
- Which data can be used, and in which environment?
- Which checks establish that the result is usable?
- Who approves the result and handles exceptions?